Letter pubs.acs.org/JPCL
CO Oxidation at the Au−Cu Interface of Bimetallic Nanoclusters Supported on CeO2(111) Liang Zhang, Hyun You Kim,*,† and Graeme Henkelman* Department of Chemistry and Biochemistry, University of Texas at Austin, Austin, Texas 78712-0165, United States S Supporting Information *
ABSTRACT: DFT+U calculations of the structure of CeO2(111)-supported Aubased bimetallic nanoclusters (NCs) show that a strong support−metal interaction induces a preferential segregation of the more reactive element to the NC−CeO2 perimeter, generating an interface with the Au component. We studied several Au -based bimetallic NCs (Au-X, X: Ag, Cu, Pd, Pt, Rh, and Ru) and found that (Au− Cu)/CeO2 is optimal for catalyzing CO oxidation via a bifunctional mechanism. O2 preferentially binds to the Cu-rich sites, whereas CO binds to the Au-rich sites. Engineering a two-component system in which the reactants do not compete for binding sites is the key to the high catalytic activity at the interface between the components. SECTION: Surfaces, Interfaces, Porous Materials, and Catalysis
T
We suggest a strategy to improve the catalytic activity of Au NPs/NCs by more intensive interface engineering, utilizing the strong metal−support interaction. We study a set of CeO2(111) supported Au-based bimetallic NCs composed of 10 atoms (Au7-X3, where X is Ag, Cu, Pd, Pt, Rh, or Ru) and find that a strong oxygen affinity of the alloying elements, X, drives their preferential segregation to the NC-CeO2(111) perimeter. Segregation of the metal components results in three interfaces between Au, CeO2, and the alloying element. CO oxidation at these interfacial sites is examined using density functional theory (DFT). The different alloying elements change the reaction energetics; Cu is found to produce a particularly active CO oxidation mechanism at the interface with Au. A 4 × 4 CeO2(111) slab model with six atomic layers and 20 Å of vacuum was prepared to describe the CeO2 support. Sensitivity tests on the model parameters (energy cutoff, kpoint sampling, and system size) showed that our calculation parameters sufficiently describe the energetics of the oxidation catalysis by CeO2-supported Au NP/NCs.1,2,14 A highly symmetric hexagonal two-layered Au NC composed of 10 atoms was supported on the CeO2(111) surface (Figure 1a). The entire Au/CeO2 system was fully optimized prior to catalysis studies. To study the effect of the CeO2(111) support on the structure of supported Au-X bimetallic NPs, we replaced three Au atoms in the top layer of the Au10 NC with one of the following alloying elements: Ag, Cu, Pd, Pt, Rh, or Ru (Figure 1b). The preferred geometry for the alloying elements in the clusters was determined using two metrics. First, the exchange
he critical role of the interface between a supporting oxide and supported metal nanoparticles (NPs)/nanoclusters (NCs) has been highlighted by many experimental and theoretical studies.1−14 Moreover, recent studies are suggesting that interfaces in nanocatalysts can be designed on the atomic scale for specific purposes. The Rodriguez and Adzic groups, in particular, have reported various kinds of tunable interfaces, metal and oxide,9,12,15−18 metal and carbide,19 and oxide and oxide,2,12,17,20 and highlighted the important role of these interfaces for various heterogeneous catalytic reactions. Since Haruta’s pioneering finding on the excellent catalytic activity of oxide-supported Au NPs, the oxidation chemistry of oxide-supported Au NPs or NCs has been studied extensively, with a focus on a determination of the active site.1,2,5,7,8 Of particular interest is how the system is able to activate the oxygen molecule to give the high catalytic activity observed experimentally. Theoretical studies of O2 activation by supported or unsupported Au NPs/NCs have reported low O2 binding energies and high O2 dissociation barriers.21 In our previous study of CO oxidation by CeO2-supported Au NCs (Au/CeO2), we found a relatively strong CO binding as compared with O2 on the Au NC of Au/CeO2(111), leading to CO poisoning and a low oxidation rate. These results suggest that a different reaction mechanism is available that involves another source of oxygen.2 In this regard, the oxygen spillover mechanism,1,2,6,14 the Mars−van Krevelen (M-vK) mechanism of CO oxidation,1,14 and O2 binding at the Au−support interface2,4 are considered as better alternatives to explain the rich chemistry of CO oxidation by oxide-supported Au catalysts that is observed experimentally. We have previously reported that the low-coordinated interfacial oxygen atoms oxidize CO bound to Au NCs (Au-CO*) by the M-vK mechanism of CO oxidation, emphasizing the role of the NC−CeO2 interface.14 © XXXX American Chemical Society
Received: July 18, 2013 Accepted: August 10, 2013
2943
dx.doi.org/10.1021/jz401524d | J. Phys. Chem. Lett. 2013, 4, 2943−2947
The Journal of Physical Chemistry Letters
Letter
A strong interaction between the CeO2 support and supported NPs/NCs and especially on the defective or stepped CeO2 surfaces has been previously reported.2,14,22,23 In the case of Au NPs/NCs, a strong interaction with the support is advantageous because pinned Au NPs/NCs are less susceptible to deactivation due to thermal sintering. Adding small amounts of an oxophilic alloying element to Au NPs/NCs can generate a pinning site of Au NPs/NCs on the Au NPs/NC perimeter and increase the lifetime of the catalyst. In a previous study of CO oxidation on the Au NC of the Au/CeO2 by the Langmuir−Hinshelwood mechanism it was found that even though the activation energies of CO oxidation by the Au−O 2 * and Au−O* are both very low, an asymmetrically strong CO binding (Ead = −1.05 eV) on the Au NC hinders the coadsorption of O2 (Ead = −0.56 eV), leading to a low O2 surface concentration and a low reaction rate.2 In the case of this Au10/CeO2 model catalyst we confirmed again that the Au3 top layer preferentially binds CO over O2 so that this surface acts as a source of bound CO molecules. However, in the case of (Au-X)/CeO2, we hypothesize that the oxophilic alloying element would generate a potential oxygen binding site, resulting in a different catalytic behavior as compared with the monometallic supported Au cluster. To validate our hypothesis, we calculated the binding energy of reactants (CO and O2) on the Au−X and X−X sites of the (Au−Cu)/CeO2, (Au−Rh)/CeO2, and (Au−Ru)/CeO2 catalysts, where the alloying element is segregated to the NC− CeO2 interface. Table 2 shows the energy of CO and O2 binding on the Au-X and X-X sites of tested catalysts and corresponding surface
Figure 1. Au10 (a) and Au7X3 (b,c) clusters supported on the CeO2(111) surface. Yellow, ivory, and red spheres represent Au, Ce, and O atoms, respectively. Green spheres in panels b and c represent the initial and final location of alloying element. Energetics associated with the segregation of alloying element to the NC-CeO2 interface (from b to c) are shown in Table 1.
energy Eex,CeO2 is calculated as the energy gained by exchanging three alloying atoms at the top layer of the (Au7-X3)/CeO2 with bottom-layer Au atoms, as shown in Figure 1c. A second measure is the CeO2-induced preferential segregation energy, Eseg,CeO2‑Au, defined as the change in exchange energy between a supported cluster and a gas phase cluster, Eseg,CeO2‑Au = Eex,CeO2 − Eex,Au, where Eex,Au is the energy gained by exchanging the three alloy atoms on the surface of a gas-phase Au cluster to the subsurface. (See Figure S1 in the Supporting Information for details.) As such, Eseg,CeO2‑Au indicates whether the bond energy between the X and the CeO2 surface is stronger than that to Au. Calculations of CO and O2 adsorption as well as the subsequent CO oxidation catalysis were examined on the most stable cluster. Figure 1 shows the structure of our models: Au/CeO2 and (Au-X)/CeO2. We have previously reported that the bonding between the CeO2 support and the supporting Au NC is governed by the hybridization of Au-5d and O-2p orbitals.14 The same nature of bonding between the Au-X and CeO2 support was found here. (See Figure S2 in the Supporting Information). The calculated values of Eseg,CeO2‑Au in the Au-X systems show that CeO2 prefers to bond with the alloying element rather than with Au atoms, showing that the CeO2 support induces a preferential segregation of the oxophilic element to the (Au-X)CeO2 perimeter. (See Table 1.) The effect of the CeO2 support on the preferential segregation of the alloying element to the NC−CeO2 interface is more prominent in the system where Eseg,CeO2‑Au is greater than Eex,Au, including Au−Cu and Au−Rh NCs. These systems, as well as the Au−Ru NC, whose Eseg,CeO2‑Au is comparable to Au−Rh, were considered for further CO oxidation studies In Au−Ag, Au−Pd, and Au−Pt NCs, the energy acquired from the CeO2-X bond formation, Eseg,CeO2‑Au, is smaller than the energy gained from the surface energy reduction, Eex,Au, indicating that the CeO2-X bond formation is not strong enough to induce a segregation of the alloying element to the NC−CeO2 interface. In larger NPs, therefore, the alloying element is expected to be found in the core of the Au-X NP rather than at the NP−CeO2 interface.
Table 2. Binding Energy and Surface Concentration of CO and O2 of Studied (Au-X)/CeO2 Catalystsa,b binding energy (eV) Au−X
O2 CO
X−X
O2 CO
Au−Cu
Au−Rh
Au−Ru
−0.64 (0.997) −0.57 (0.003) −0.77 (1.00) −0.60 (0.00)
−1.02 (0.00) −1.37 (1.00) −1.57 (0.007) −1.78 (0.993)
−1.19 (0.00) −1.64 (1.00) −1.31 (0.00) −1.77 (1.00)
a Binding sites are shown in Figure 2. bValues in the parentheses represent the surface concentration of the corresponding reactant at 298 K, p(O2) = 0.21, and p(CO) = 0.01. (See the Supporting Information for details.)
concentration of CO and O2 at the binding site. (The Supporting Information has details of the surface concentration calculations.) The binding sites in the Au−Ru and Au−Rh systems that bind CO more strongly than O2 would be saturated by CO when the catalyst is exposed to the CO oxidation condition. (See Table 2.) Because these systems do not have a preferential O2 binding site and their CO binding
Table 1. Calculated values of Eex,CeO2, Eex,Au, and Eseg,CeO2‑Au alloying element
Ag
Cu
Pd
Pt
Rh
Ru
Eex,CeO2 (eV)
−0.89
−2.56
−1.30
−1.25
−2.63
−3.69
Eex,Au (eV) Eseg,CeO2‑Au (eV)
−0.74 −0.15
−1.17 −1.39
−0.97 −0.33
−0.94 −0.31
−0.98 −1.65
−2.08 −1.61
2944
dx.doi.org/10.1021/jz401524d | J. Phys. Chem. Lett. 2013, 4, 2943−2947
The Journal of Physical Chemistry Letters
Letter
Figure 2. Two available initial CO oxidation channels catalyzed by Au7Cu3 NC supported on CeO2(111). The ET channel (b) provides the faster pathway with a lower activation energy (ΔETS). Ivory, red, blue, gray, and green spheres represents Ce, O (CeO2), O (O2), C, and O(CO) atoms, respectively. Au and Cu atoms in the Au7Cu3 NC were colored in yellow and copper. ΔEx is the energy of the xth state relative to the previous stage; for example, ΔE2 is the energy difference between stage 2 and stage 1.
Benefiting from its ability to separate the CO and O2 binding sites, the (Au−Cu)/CeO2 system was selected for analysis of its catalytic activity. Coadsorption of CO and O2 was tested on the region of Au−Cu interface. As shown in Figure 2, the top Au sites are dominated by CO, and O2 can binds to either bottom Cu−Cu sites or edge Au−Cu sites in a bridge geometry, initiating two possible reaction pathways. These two reaction pathways are denoted as (1) BT: where O2 binds to the Cu− Cu sites in a bridge geometry and CO binds to the top layer Au atom and (2) ET: where O2 binds to the Au−Cu sites in a bridge geometry and CO binds to the top layer Au atom. Figure 2 shows the overall energy profiles of two reaction pathways. The coadsorption geometry of both reaction channels is almost equally favored: ΔE1 (BT) = −1.50 eV and ΔE1 (ET) = −1.39 eV. Association of coadsorbed CO and O2, which is the ratedetermining step of CO oxidation by the Langmuir−Hinshelwood reaction,2,26 produces a gas phase CO2 and a residual Au−O* with an activation energy of 0.53 eV for the BT channel and 0.11 eV for the ET channel. The accessible O2 adsorption geometry in the ET channel (one of the O atoms of the adsorbed O2 molecule is close to the Au-CO* species) lowers the activation energy of the first CO oxidation step, making the ET channel the favored CO oxidation pathway. After the first CO oxidation step, the residual Au−O* oxidized one more CO molecule, completing the CO oxidation process. Figure 3 shows the energy profile of the second CO oxidation by the Au−O*, which proceeds with an activation barrier of 0.23 eV. We have reported atomic oxygen on the Au NCs/NPs as a highly reactive species.2 In the case of the Au−Cu bimetallic NC, however, additional stabilizing effect from the oxophilic Cu atoms likely contributes to the increased barrier of CO oxidation by the Au−O* binding energy. We speculate that this barrier would increase as a function of the Cu concentration in the Au−Cu bimetallic NP. At higher Cu concentrations, oxygen atoms would oxidize Cu atoms, converting them to Cu2O or CuO2 deactivating the CO oxidation mechanism reported here.
energy is higher than that of the Au/CeO2 system, CO poisoning at the surface of Au−Rh and Au−Ru NCs would prevent CO oxidation by the Langmuir−Hinshelwood mechanism. We should note here that DFT at the GGA level of theory is known to have systematic errors in the binding energy of molecules, arising, for example, from the reference energy of gas-phase O2. Thus, the relative binding energies and reaction rates between different catalysts (as reported in Table 2) should be trusted more than the absolute values. An oxygen spillover mechanism, involving the diffusion of a lattice oxygen atom of the CeO2 support to the supported oxophilic metal NPs/NCs, was found to be endothermic in the Au−Rh and Au−Ru systems; 0.77 and 0.63 eV in the (Au− Rh)/CeO2 and (Au−Ru)/CeO2, respectively. Even the O2 binding energy in the Au−Rh and Au−Ru systems is stronger than that in Au−Cu; their absolute O2 binding energy is far less than the vacancy formation energy of the CeO2(111) surface, which is 2.48 eV in our system. Therefore, there is no mechanism to provide a sufficient concentration of the (Au− Rh)-O* and (Au−Ru)-O* species for CO oxidation. Some caution of this result is appropriate, however, because oxygen spillover from the CeO2 support to the supported pure Ag24 and Pt6 NCs/NPs has been reported. The case of Ag NPs is still controversial; Luches et al. claimed that oxygen spillover cannot occur from the CeO2(111) surface.25 For Pt NP, oxygen spillover was reported from the low-coordinated oxygen atom of nanosized CeO2. Modifying the oxygen binding energy to the NC by increasing the concentration of the oxophilic alloying element or decreasing the vacancy formation energy of the CeO2 support by reducing the size may facilitate oxygen spillover and subsequent CO oxidation. The (Au−Cu)/CeO2 is the only system where both Au−Cu and Cu−Cu sites clearly prefer to bind O2 more strongly than CO. The binding sites of O2 and CO are well-separated in this system; that is, O2 binds to the alloying atom sites, while CO binds to the Au sites. 2945
dx.doi.org/10.1021/jz401524d | J. Phys. Chem. Lett. 2013, 4, 2943−2947
The Journal of Physical Chemistry Letters
Letter
point. The convergence criteria for the electronic structure and the atomic geometry were 10−4 eV and 0.01 eV/Å, respectively. Sensitivity tests show that our results are robust with respect to the calculation and model parameters, including the k-point grid, cutoff energy, and thickness of the slab.1,2,14 Increasing the energy cutoff to 500 eV changed the coadsorption energy of CO and O2 on the segregated CeO2/AuCu cluster (considering both coadsorption geometries shown in Figure 2) by